In common techniques, measurement agents placed on a network measure qualities of services provided by provision servers, and a surveillance server analyzes results of measurement, thereby specifying a failure location (for example, a faulty router or switch, or a router or switch in which a packet transfer process is delayed).
A technique that enables measurement agents to inform a surveillance server of all information measured by the measurement agents and enables the surveillance server to specify a failure location is known as one of the common techniques (see Japanese Laid-open Patent Publication No. 2000-242585). Specifically, as depicted in FIG. 28, measurement agents placed on a network measure reception qualities and transmission paths of the services, and the surveillance server analyzes all measurement results. The surveillance server analyzes the measurement results received from the measurement agents and specifies a portion through which many of the services with deteriorated qualities pass as a cause of a failure.
Japanese Laid-open Patent Publication No. 11-275106 discloses a technique that enables to group a plurality of communicating devices located within a communicable range and to regard one of the communicating devices as a server and the others as clients. It is also possible to group measurement agents by using this technique and notify a surveillance server of measurement results in units of groups.
However, in the technique described in Japanese Laid-open Patent Publication No. 2000-242585, the surveillance server receives all the measurement results and analyzes the measurement results. Accordingly, the surveillance server manages a large amount of information, which causes memory overflow or the like and consequently prevents quick detection of a failure.
Further, in the technique using the technique described in Japanese Laid-open Patent Publication No. 11-275106, only the group of the measurement agents located within the communicable range is determined. Accordingly, loads on the surveillance server are great as in the technique above mentioned, which causes memory overflow or the like and consequently prevents quick failure detection.